annotate validate_affy_metadata.py @ 25:d9f3bcfeecfe draft default tip

Uploaded
author greg
date Thu, 15 Aug 2019 13:15:22 -0400
parents 8a826d1afe69
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
1 #!/usr/bin/env python
ae030e6e838c Uploaded
greg
parents:
diff changeset
2 """
ae030e6e838c Uploaded
greg
parents:
diff changeset
3 Validate the metadata file associated with Affymetrix 96 well plate data.
ae030e6e838c Uploaded
greg
parents:
diff changeset
4 """
ae030e6e838c Uploaded
greg
parents:
diff changeset
5 import argparse
ae030e6e838c Uploaded
greg
parents:
diff changeset
6 import datetime
ae030e6e838c Uploaded
greg
parents:
diff changeset
7 import decimal
ae030e6e838c Uploaded
greg
parents:
diff changeset
8 import re
ae030e6e838c Uploaded
greg
parents:
diff changeset
9 import shutil
ae030e6e838c Uploaded
greg
parents:
diff changeset
10 import sys
ae030e6e838c Uploaded
greg
parents:
diff changeset
11
ae030e6e838c Uploaded
greg
parents:
diff changeset
12 parser = argparse.ArgumentParser()
ae030e6e838c Uploaded
greg
parents:
diff changeset
13 parser.add_argument('--input', dest='input', help='Metadata file for Affymetrix 96 well plate data')
ae030e6e838c Uploaded
greg
parents:
diff changeset
14 parser.add_argument('--output', dest='output', help='Output dataset'),
ae030e6e838c Uploaded
greg
parents:
diff changeset
15 args = parser.parse_args()
ae030e6e838c Uploaded
greg
parents:
diff changeset
16
ae030e6e838c Uploaded
greg
parents:
diff changeset
17 EMAIL_MAX_LEN = 255
ae030e6e838c Uploaded
greg
parents:
diff changeset
18 VALID_EMAIL_RE = re.compile("[^@]+@[^@]+\.[^@]+")
ae030e6e838c Uploaded
greg
parents:
diff changeset
19
ae030e6e838c Uploaded
greg
parents:
diff changeset
20
ae030e6e838c Uploaded
greg
parents:
diff changeset
21 def add_error_msg(accumulated_msgs, msg):
ae030e6e838c Uploaded
greg
parents:
diff changeset
22 return "%s\n%s" % (accumulated_msgs, msg)
ae030e6e838c Uploaded
greg
parents:
diff changeset
23
ae030e6e838c Uploaded
greg
parents:
diff changeset
24
ae030e6e838c Uploaded
greg
parents:
diff changeset
25 def empty_value(line_no, label, accumulated_msgs):
ae030e6e838c Uploaded
greg
parents:
diff changeset
26 return add_error_msg(accumulated_msgs, "The required %s value is missing on line %d." % (label, line_no))
ae030e6e838c Uploaded
greg
parents:
diff changeset
27
ae030e6e838c Uploaded
greg
parents:
diff changeset
28
ae030e6e838c Uploaded
greg
parents:
diff changeset
29 def stop_error(msg):
ae030e6e838c Uploaded
greg
parents:
diff changeset
30 sys.exit(msg)
ae030e6e838c Uploaded
greg
parents:
diff changeset
31
ae030e6e838c Uploaded
greg
parents:
diff changeset
32
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
33 def string_as_boolean_string(string):
d15ae203013e Uploaded
greg
parents: 8
diff changeset
34 if str(string).lower() in ['true', 'yes', 'on', '1']:
d15ae203013e Uploaded
greg
parents: 8
diff changeset
35 return 'True'
d15ae203013e Uploaded
greg
parents: 8
diff changeset
36 else:
d15ae203013e Uploaded
greg
parents: 8
diff changeset
37 return 'False'
d15ae203013e Uploaded
greg
parents: 8
diff changeset
38
d15ae203013e Uploaded
greg
parents: 8
diff changeset
39
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
40 def validate_date_string(line_no, date_string, column, accumulated_msgs):
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
41 if len(date_string) == 0:
d15ae203013e Uploaded
greg
parents: 8
diff changeset
42 return accumulated_msgs
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
43 try:
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
44 datetime.datetime.strptime(date_string, '%Y-%m-%d')
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
45 return accumulated_msgs
ae030e6e838c Uploaded
greg
parents:
diff changeset
46 except ValueError:
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
47 return add_error_msg(accumulated_msgs, "Line %d contains an incorrect date format (%s must be YYYY-MM-DD) for column %s." % (line_no, date_string, column))
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
48
ae030e6e838c Uploaded
greg
parents:
diff changeset
49
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
50 def validate_decimal(line_no, decimal_string, column, accumulated_msgs):
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
51 try:
ae030e6e838c Uploaded
greg
parents:
diff changeset
52 decimal.Decimal(decimal_string)
ae030e6e838c Uploaded
greg
parents:
diff changeset
53 return accumulated_msgs
ae030e6e838c Uploaded
greg
parents:
diff changeset
54 except Exception:
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
55 return add_error_msg(accumulated_msgs, "Line %d contains an incorrect decimal value (%s) for column %s." % (line_no, decimal_string, column))
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
56
ae030e6e838c Uploaded
greg
parents:
diff changeset
57
ae030e6e838c Uploaded
greg
parents:
diff changeset
58 def validate_email(line_no, email, accumulated_msgs):
ae030e6e838c Uploaded
greg
parents:
diff changeset
59 if not (VALID_EMAIL_RE.match(email)):
ae030e6e838c Uploaded
greg
parents:
diff changeset
60 return add_error_msg(accumulated_msgs, "Line %d contains an invalid email address (%s). " % (line_no, email))
ae030e6e838c Uploaded
greg
parents:
diff changeset
61 elif len(email) > EMAIL_MAX_LEN:
ae030e6e838c Uploaded
greg
parents:
diff changeset
62 return add_error_msg(accumulated_msgs, "Line %d contains an email address (%) that is longer than the maximum length, %d characters." % (line_no, email))
ae030e6e838c Uploaded
greg
parents:
diff changeset
63 return accumulated_msgs
ae030e6e838c Uploaded
greg
parents:
diff changeset
64
ae030e6e838c Uploaded
greg
parents:
diff changeset
65
ae030e6e838c Uploaded
greg
parents:
diff changeset
66 accumulated_msgs = ""
ae030e6e838c Uploaded
greg
parents:
diff changeset
67 # Parse the input file, skipping the header, and validating
ae030e6e838c Uploaded
greg
parents:
diff changeset
68 # that each data line consists of 31 comma-separated items.
ae030e6e838c Uploaded
greg
parents:
diff changeset
69 with open(args.input, "r") as ih:
1
613186952d42 Uploaded
greg
parents: 0
diff changeset
70 for i, line in enumerate(ih):
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
71 if i == 0:
ae030e6e838c Uploaded
greg
parents:
diff changeset
72 # Skip the header.
ae030e6e838c Uploaded
greg
parents:
diff changeset
73 continue
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
74 # Keep 1-based line value for error messages.
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
75 line_no = i + 1
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
76 line = line.rstrip("\r\n")
ae030e6e838c Uploaded
greg
parents:
diff changeset
77 if i > 97:
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
78 accumulated_msgs = add_error_msg(accumulated_msgs, "The input file contains more than 97 lines (must be 1 header line and no more than 96 data lines).")
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
79 stop_error(accumulated_msgs)
17
17db886fbfad Uploaded
greg
parents: 13
diff changeset
80 items = line.split("\t")
23
84478b80cc47 Uploaded
greg
parents: 21
diff changeset
81 if len(items) != 32:
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
82 accumulated_msgs = add_error_msg(accumulated_msgs, "Line %d contains %s columns, (must be 32)." % (line_no, len(items)))
2
f3ca1cd544fa Uploaded
greg
parents: 1
diff changeset
83 stop_error(accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
84 # Required and validated.
ae030e6e838c Uploaded
greg
parents:
diff changeset
85 # Required.
18
38166b76f47a Uploaded
greg
parents: 17
diff changeset
86 user_specimen_id = items[0]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
87 if len(user_specimen_id) == 0:
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
88 accumulated_msgs = empty_value(line_no, "user_specimen_id", accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
89 # Optional.
18
38166b76f47a Uploaded
greg
parents: 17
diff changeset
90 field_call = items[1]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
91 # Optional.
18
38166b76f47a Uploaded
greg
parents: 17
diff changeset
92 bcoral_genet_id = items[2]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
93 # Optional.
18
38166b76f47a Uploaded
greg
parents: 17
diff changeset
94 bsym_genet_id = items[3]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
95 # Required.
18
38166b76f47a Uploaded
greg
parents: 17
diff changeset
96 reef = items[4]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
97 if len(reef) == 0:
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
98 accumulated_msgs = empty_value(line_no, "reef", accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
99 # Required.
18
38166b76f47a Uploaded
greg
parents: 17
diff changeset
100 region = items[5]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
101 if len(region) == 0:
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
102 accumulated_msgs = empty_value(line_no, "region", accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
103 # Required and validated.
18
38166b76f47a Uploaded
greg
parents: 17
diff changeset
104 latitude = items[6]
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
105 accumulated_msgs = validate_decimal(line_no, latitude, "latitude", accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
106 # Required and validated.
18
38166b76f47a Uploaded
greg
parents: 17
diff changeset
107 longitude = items[7]
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
108 accumulated_msgs = validate_decimal(line_no, longitude, "longitude", accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
109 # Optional.
18
38166b76f47a Uploaded
greg
parents: 17
diff changeset
110 geographic_origin = items[8]
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
111 # Optional.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
112 colony_location = items[9]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
113 # Optional.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
114 depth = items[10]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
115 # Optional.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
116 disease_resist = items[11]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
117 # Optional.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
118 bleach_resist = items[12]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
119 # Optional.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
120 mortality = items[13]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
121 # Optional.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
122 tle = items[14]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
123 # Optional.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
124 spawning = string_as_boolean_string(items[15])
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
125 # Required.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
126 collector_last_name = items[16]
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
127 if len(collector_last_name) == 0:
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
128 accumulated_msgs = empty_value(line_no, "collector_last_name", accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
129 # Required.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
130 collector_first_name = items[17]
9
d15ae203013e Uploaded
greg
parents: 8
diff changeset
131 if len(collector_first_name) == 0:
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
132 accumulated_msgs = empty_value(line_no, "collector_first_name", accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
133 # Required.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
134 org = items[18]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
135 if len(org) == 0:
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
136 accumulated_msgs = empty_value(line_no, "org", accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
137 # Required and validated.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
138 collection_date = items[19]
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
139 accumulated_msgs = validate_date_string(line_no, collection_date, "collection_date", accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
140 # Required and validated.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
141 contact_email = items[20]
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
142 accumulated_msgs = validate_email(line_no, contact_email, accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
143 # Required.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
144 seq_facility = items[21]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
145 if len(seq_facility) == 0:
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
146 accumulated_msgs = empty_value(line_no, "seq_facility", accumulated_msgs)
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
147 # Optional.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
148 array_version = items[22]
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
149 # Optional.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
150 public = string_as_boolean_string(items[23])
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
151 # Optional.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
152 public_after_date = items[24]
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
153 accumulated_msga = validate_date_string(line_no, public_after_date, "public_after_date", accumulated_msgs)
21
90688364136c Uploaded
greg
parents: 18
diff changeset
154 # Required and validated.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
155 sperm_motility = items[25]
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
156 accumulated_msgs = validate_decimal(line_no, sperm_motility, "sperm_motility", accumulated_msgs)
21
90688364136c Uploaded
greg
parents: 18
diff changeset
157 # Required and validated.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
158 healing_time = items[26]
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
159 accumulated_msgs = validate_decimal(line_no, healing_time, "healing_time", accumulated_msgs)
23
84478b80cc47 Uploaded
greg
parents: 21
diff changeset
160 # Optional.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
161 dna_extraction_method = items[27]
23
84478b80cc47 Uploaded
greg
parents: 21
diff changeset
162 # Optional.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
163 dna_concentration = items[28]
23
84478b80cc47 Uploaded
greg
parents: 21
diff changeset
164 # If dna_concentration has a value, then it must be decimal.
84478b80cc47 Uploaded
greg
parents: 21
diff changeset
165 if len(dna_concentration) > 0:
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
166 accumulated_msgs = validate_decimal(line_no, dna_concentration, "dna_concentration", accumulated_msgs)
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
167 # Optional.
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
168 registry_id = items[29]
24
8a826d1afe69 Uploaded
greg
parents: 23
diff changeset
169 # Optional.
25
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
170 result_folder_name = items[30]
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
171 # Optional.
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
172 plate_barcode = items[31]
d9f3bcfeecfe Uploaded
greg
parents: 24
diff changeset
173
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
174
6
11d22a27c24a Uploaded
greg
parents: 5
diff changeset
175 if len(accumulated_msgs) > 0:
0
ae030e6e838c Uploaded
greg
parents:
diff changeset
176 stop_error(accumulated_msgs)
ae030e6e838c Uploaded
greg
parents:
diff changeset
177
8
0735cfcc7490 Uploaded
greg
parents: 7
diff changeset
178 shutil.copyfile(args.input, args.output)